Sub-pixel motion estimation using gradient cross-correlation

A highly accurate and computationally efficient method is presented suitable for the estimation of motion in video sequences. The method is based on the maximisation of the spatial gradient cross-correlation function, which is computed in the frequency domain and therefore is implemented by fast transformation algorithms. By taking into full consideration gradient magnitude and phase, the selection of salient and reliable image features is promoted and the resulting accuracy of motion estimation enhanced. The proposed method outperforms established competing frequency-domain motion estimation methods, most notably phase correlation, in terms of sub-pixel accuracy for a range of test material and motion scenarios. Our results also show that our method is considerably more immune to the presence of manually induced additive Gaussian noise.